Table 2 NLP and ICD10CM validation.

From: Improving ascertainment of suicidal ideation and suicide attempt with natural language processing

Patient selection

Outcome

P@200

AUPRC (95% CI)

A

Top 200 patients retrieved by the NLP system

SI

98.5

98.6 (97.1—99.5)

SA

96.5

97.3 (95.2—98.7)

Patient selection

Outcome

N

P

B

Patients in top 200 w/ relevant ICD codes

SI

170

100

SA

149

98.7

Patients in top 200 w/o relevant ICD codes

SI

30

90.0

SA

51

90.2

Patients in top 200 w/ 1+ positive mentions

SI

200

98.5

SA

199

97.0

C

Patients with ICD10CM codes for suicide

SI

200

96.0

SA

200

85.0

  1. AUPRC area under the precision-recall curve, CI confidence interval, P@K precision at top K retrieved patients, P precision, SI suicidal ideation, SA suicide attempt.